Learning Whole-Body Human-Humanoid Interaction from Human-Human Demonstrations

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• Standard retargeting methods fail to generate usable Human-Humanoid Interaction data because they break essential contacts between participants.limitation 90%
  Evidencepartial

  we first demonstrate that standard retargeting fails by breaking the essential contacts

  Implicationpartial

  Directly stated in the abstract as a key finding that motivates the research

  Verificationpartial

  partial

• PAIR (Physics-Aware Interaction Retargeting) generates physically consistent HHoI data by preserving contact semantics across morphology differences.method 90%
  Evidencepartial

  PAIR (Physics-Aware Interaction Retargeting), a contact-centric, two-stage pipeline that preserves contact semantics across morphology differences to generate physically consistent HHoI data

  Implicationpartial

  Explicitly described as the proposed solution with clear methodology

  Verificationpartial

  partial

• Conventional imitation learning policies trained on high-quality HHoI data fail because they merely mimic trajectories and lack interactive understanding.limitation 85%
  Evidencepartial

  conventional imitation learning policies merely mimic trajectories and lack interactive understanding

  Implicationpartial

  Directly stated as a second failure mode discovered when using the generated data

  Verificationpartial

  partial

• D-STAR (Decoupled Spatio-Temporal Action Reasoner) uses hierarchical policy design to disentangle when to act from where to act.method 90%
  Evidencepartial

  D-STAR (Decoupled Spatio-Temporal Action Reasoner), a hierarchical policy that disentangles when to act from where to act

  Implicationpartial

  Explicitly described as the proposed solution with clear architectural details

  Verificationpartial

  partial

• D-STAR's decoupled reasoning streams enable learning robust temporal phases without being distracted by spatial noise.technical 85%
  Evidencepartial

  By decoupling these reasoning streams, our model learns robust temporal phases without being distracted by spatial noise

  Implicationpartial

  Directly stated as a benefit of the architectural design

  Verificationpartial

  partial

• The proposed framework demonstrates significant performance gains over baseline approaches in extensive simulations.result 85%
  Evidencepartial

  We validate our framework through extensive and rigorous simulations, demonstrating significant performance gains over baseline approaches

  Implicationpartial

  Directly stated validation results with clear comparative claims

  Verificationpartial

  partial

• The complete pipeline enables learning complex whole-body interactions from Human-Human Interaction data.result 80%
  Evidencepartial

  a complete, effective pipeline for learning complex whole-body interactions from HHI data

  Implicationpartial

  Directly stated as the overall contribution of the work

  Verificationpartial

  partial

• Scarcity of high-quality Human-Humanoid Interaction data hinders progress in enabling humanoid robots to physically interact with humans.limitation 90%
  Evidencepartial

  progress is hindered by the scarcity of high-quality Human-Humanoid Interaction (HHoI) data

  Implicationpartial

  Directly stated as the fundamental problem motivating the research

  Verificationpartial

  partial

Startup potential card